Pretty fun. Didn't expect to be participating in some kind of meta-level prisoner's dilemma with the presentation itself hehe. I had to adjust my expectations to predict a total of 3/4. I still want to know what the overall research results tell us about LLMs. It sounds like they don't do tit-for-tat, and try to defect more than humans.
Pretty fun. Didn't expect to be participating in some kind of meta-level prisoner's dilemma with the presentation itself hehe. I had to adjust my expectations to predict a total of 3/4. I still want to know what the overall research results tell us about LLMs. It sounds like they don't do tit-for-tat, and try to defect more than humans.
I am still doing the research, but my feeling is that LLMs play tit for tat - if you don't provoke them, they are told to cooperate
Here are some initial results -
https://thenunchiproject.com/blog/ai-betrays-you-in-the-last...
I am glad you enjoyed the experience - this is my first time posting on HN so this made my day :)
[flagged]
[flagged]